Screen failure rate calculator: how many to screen for your target

The arithmetic is trivial. The disagreement is almost always about the two inputs. A screen failure is commonly a person who signed consent and started screening procedures but was not randomised or enrolled, whether because a criterion was not met, a test came back out of range, or they withdrew before the end of screening. People who were only pre-screened and never consented are usually not counted, because they never entered the study record. Your protocol may define it differently, and that definition wins. The point is to pick one, put it in the statistical or operations plan, and apply it at every site.

The denominator needs the same discipline. If someone is screened twice, are they one screened participant or two? If a participant fails, is re-screened and then randomised, do you count one failure and one success, or only the success? A screening and enrollment log that records every screening attempt with a reason for each failure makes these choices explicit. Whatever you decide, apply it consistently, because the same data can yield a rate of 30.0% or 32.3% depending on whether 10 re-screens are counted as extra screenings (42 failures over 140 attempts, or over 130 distinct people).

A rate observed on 20 or 60 screened people is a noisy estimate. A simple way to see how noisy is to put an interval around it. One standard choice for a proportion is the Wilson score interval. For the same observed 30%, the 95% interval is wide at small numbers and narrows as screenings accumulate: 6 of 20 gives roughly 15% to 52%; 18 of 60 gives roughly 20% to 43%; 24 of 80 gives roughly 21% to 41%; 180 of 600 gives roughly 26% to 34%.

Use this when reviewing early data. If your plan assumed 20% and your first 60 screenings show 30% with an interval of 20% to 43%, the plan is not yet contradicted, but the upper end deserves a contingency: at 43% you would need 120 / 0.57 = 211 screenings, not 150. A statistician may prefer a different interval method, such as an exact one; the principle is the same. Report the count alongside the percentage so that readers can judge.

The rate tells you how many, the reasons tell you why. Group failures by criterion. In most programmes a small number of criteria account for most failures, often a laboratory threshold, a washout period or a comorbidity that is more common than the protocol writers assumed. If one exclusion criterion accounts for a large share of failures, ask three questions: could sites check it earlier through pre-screening, is the threshold clinically necessary, and is it worded so that sites apply it the same way? Changing a criterion needs a protocol amendment and ethics approval, so find out early rather than late.

Two other levers are cheaper. First, move the high-failure checks earlier in the visit so that fewer procedures are done on people who will fail. Second, make the criteria testable at pre-screening from a medical record, which keeps non-consented failures out of the study denominator and out of the burden on participants. To turn the rate and the screening pace into a calendar date, use the enrollment forecast calculator; to write the plan around it, use the recruitment plan template.